1. Trajectory analysis of pointing movements: how many trials are needed for reliable data?2
Jarrod Blinch1, Youngdeok Kim1, Romeo Chua2 1
motorbehaviour.wordpress.com
Introduction
Method, continued
Results, continued
Displacement profile (%)
M ± SE (mm)
D study
g ≥ .80 10
5 ± 0.7 3 20
20 ± 1.3 5 30
44 ± 1.8 6 40
75 ± 2.3 7 50
108 ± 2.5 60
139 ± 2.3 8 70
165 ± 1.8 9 80
184 ± 1.3 12 90
197 ± 0.7 23
100
202 ± 0.5 47
A powerful tool in motor behaviour research is trajectory analysis. A vital issue that requires further investigation is how many trials are needed for a reliable measure of trajectories. The purpose of the present study was to estimate the minimum number of trials required to achieve the conventional level of reliability for trajectory analysis. We analysed the following basic measurements of movement and two common methods of trajectory analysis within the framework of generalisability theory.
We examined kinematic landmarks throughout movement execution; specifically, time and magnitude of positive peak acceleration, peak velocity, and negative peak acceleration (shown below). For the analysis of the time-normalised displacement profiles, we examined the movements from 10 to 100% time in increments of 10%.
Basic measurement of movement
Trajectory analysis
Reaction time
Kinematic landmarks
Movement time
Time-normalised displacement profiles
Constant error
We used generalisability theory to estimate the variance contributed by the following sources of measurement error.
Results
Basic measurement of movement
M ± SE
D study
g ≥ .80
Reaction time
Choice right long
284 ± 9.0 ms 6
Simple right long
262 ± 6.5 ms 10
Movement time
304 ± 12.2 ms 3
Constant error
1.94 ± 0.45 mm 47
Source
Variability
Person
Inter-individual
True variance
Trial
Inter-trial
Systematic error
Person-by-trial
Intra-individual
Random error
Conclusions
A decision study can then estimate the change in reliability by increasing and decreasing the number of trials. This is done to determine the optimal measurement conditions to achieve the conventional level of reliability [generalisability coefficient (g) ≥ .80] for each dependent variable.
The number of trials needed for reliable data depended on the dependent variable
Reaction time required six and ten trials in choice and simple reaction time tasks
Movement time required three trials, whereas constant error required 47
The time and magnitude of kinematic landmarks required five trials or less
Time-normalised displacement profiles required more trials as the movement progressed
Kinematic landmark
M ± SE
D study
g ≥ .80
Time (ms)
Positive peak acceleration
41 ± 3.8 5
Peak velocity
134 ± 5.7 3
Negative peak acceleration
248 ± 10.8
Magnitude
Positive peak acceleration (mm/s2)
14,673 ± 1,142 2
Peak velocity (mm/s)
1,174 ± 52 1
Negative peak acceleration (mm/s2)
-12,930 ± 1,586
Method
The data for the present experiment were previously reported in Blinch et al. (2014). Twenty participants made unimanual pointing movements to targets (15.3 mm radius) with long (200 mm) or short (100 mm) movement amplitudes. Left and right unimanual movements were made in simple and choice reaction time blocks. Each block included 32 test trials of each movement type. Participants pointed with a stylus and were instructed to “hit the targets as quickly as possible.” An infrared emitting diode was placed near the surface of the stylus and its position was recorded with an Optotrak (3020, Northern Digital Inc.) at 250 Hz.
2017